Um processo de engenharia de requisitos para sistemas multiagentes

Abstract

In software engineering, the requirements engineering subárea (RE) is a crucial phase for all the development process. It is not different for multiagent systems (MAS). This kind of system has demonstrated to be a good alternative to dealing with complex solutions, given its capability to divide the complexity among several agents. However, multiagent systems have specific requirements, such as beliefs, desires and intentions. This information can be represented by means of Belief-Desire-Intention (BDI) model, one of the most knew approachs to integrate the cognitive habilities desired to autonomous agents. Beyond of the requirements needed to the BDI model application, we can highlight the perceptions that an agent can have about the environment, the plans of this agent to achieve a goal (intention) and the actions that these agents can perform in the environment while trying to achieve a goal. Keeping in mind these requirements, we realized the necessity of a requirements engineering process for multiagent systems. This way, this work has as its goal to present a process supporting the RE subareas, elicitation, analysis, sp ecification and validation and allowing the representation of these specific requirements for multiagent systems. During the conception of this process, we analysed processes that support RE for MAS, performing a systematic literature review, aiming to identify the strengths and weaknesses of these processes. Thus, we defined the process phases and we proposed a RE process for MAS. The proposed process was validated by its application in Heraclito system, in which we demonstrated the execution of the process phases and results obtained in each one of them, highlighting the scenarios documentation and usecase diagrams produced during the execution. With the process application, we managed to evaluate its phases, to identify future works and improvements points in the process.